Data fusion via intrinsic dynamic variables: An application of data-driven Koopman spectral analysis

Matthew O. Williams, Clarence Worth Rowley, Igor Mezić, Yannis Kevrekidis

Research output: Contribution to journalArticlepeer-review

42 Scopus citations

Abstract

We demonstrate that the Koopman eigenfunctions and eigenvalues define a set of intrinsic coordinates, which serve as a natural framework for fusing measurements obtained from heterogeneous collections of sensors in systems governed by nonlinear evolution laws. These measurements can be nonlinear, but must, in principle, be rich enough to allow the state to be reconstructed. We approximate the associated Koopman operator using extended dynamic mode decomposition, so the method only requires time series of data for each set of measurements, and a single set of "joint" measurements, which are known to correspond to the same underlying state. We apply this procedure to the FitzHugh-Nagumo PDE, and fuse measurements taken at a single point with principal-component measurements.

Original languageEnglish (US)
Article number40007
JournalEPL
Volume109
Issue number4
DOIs
StatePublished - Feb 1 2015

All Science Journal Classification (ASJC) codes

  • General Physics and Astronomy

Fingerprint

Dive into the research topics of 'Data fusion via intrinsic dynamic variables: An application of data-driven Koopman spectral analysis'. Together they form a unique fingerprint.

Cite this